List your company
ResearchFormerly WizWand

Speech Reconstruction on English speech

3.92UTMOSWavTokenizer
1.14321.86412.5853.3059Jan 14, 2026
Updated 8mo ago

Evaluation results

MethodLinks
WavTokenizerToken rate=75 Hz, BitRate (kbps)=0.92026.013.923.2384
SACfs=12.5 Hz, fa=50 Hz, BitRate (kbps)=0.8752026.013.882.0383
DualCodecsemantic_rate=12.5 Hz, Layers=6-layer, BitRate (kbps)=0.9252026.013.782.5876
DSA-Tokenizerfs=25 Hz, fa=50 Hz, FSQ_levels_channels=4^8, Stacked_Layers=1, BitRate (kbps)=1.12026.013.675.9360
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^8, Stacked_Layers=1, BitRate (kbps)=0.72026.013.586.6757
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^7, Stacked_Layers=2, BitRate (kbps)=12026.013.568.7760
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^8, Stacked_Layers=2, BitRate (kbps)=1.12026.013.548.4161
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^8, Stacked_Layers=2, BitRate (kbps)=1.12026.013.462.6177
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^7, Stacked_Layers=2, BitRate (kbps)=12026.013.442.4576
DSA-Tokenizerfs=25 Hz, fa=25 Hz, FSQ_levels_channels=4^8, Stacked_Layers=1, BitRate (kbps)=0.72026.013.412.0972
DSA-Tokenizerfs=25 Hz, fa=50 Hz, FSQ_levels_channels=4^8, Stacked_Layers=1, BitRate (kbps)=1.12026.013.382.4976
MimiLayers=8-layer, BitRate (kbps)=1.12026.013.33.4674
DualCodecsemantic_rate=12.5 Hz, Layers=6-layer, BitRate (kbps)=0.9252026.012.3917.3611
MimiLayers=8-layer, BitRate (kbps)=1.12026.012.19107.5153
SpeechTokenizerLayers=2-layer, BitRate (kbps)=12026.012.127.5736
EncodecLayers=2-layer, BitRate (kbps)=1.52026.011.565.461
SpeechTokenizerLayers=2-layer, BitRate (kbps)=12026.011.412.9813
SACfs=12.5 Hz, fa=50 Hz, BitRate (kbps)=0.8752026.011.3490.2213
EncodecLayers=2-layer, BitRate (kbps)=1.52026.011.2598.319